Model to predict soil parent material underlying a loess mantle in illinois from satellite data1

Low accuracy of soil spectral classes within areas with thin surficial mantles is often due to confusion caused by differences in the underlying parent materials in large imageries. This study was conducted to determine whether satellite digital data could be used to predict underlying soil parent material at specific locations in a glaciated landscape with a thin loess mantle. Soils were sampled on two regular grids, each covering an area of 3,108 ha, in Ford County, Illinois. The soils developed in thin loess over both loamy and clayey glacial till, outwash, and lacustrine sediments of the Wisconsinan glaciation. Satellite spectral data obtained over parent materials underlying a thin loess mantle and from an undulating site were statistically analyzed by multipleregression procedure to develop a model (linear equation) to predict underlying soil parent material with a thin loess mantle. Correlations between underlying parent material and the red band, green band, and brightness index were significant (0.212, 0.267, and 0.176, respectively) at the 0.01 level. The absence of loess mantle at some of the undulating reference sites due to erosion and the presence of a new parent material underlying the loess mantle at the test site contributed to the differences between predicted and observed underlying parent materials. For these reasons correlations presented only accounted for a small amount of the variation. Model-predicted underlying parent materials of the nearly level test area showed a significant correlation of 0.148 with actual values. The possibility for reasonable subdivision of an image suggests a potential for improving spectral soil maps by separation of large survey areas with thin loess mantles by underlying parent materials prior to computer classification.